Smaller (datacenters) decrease...larger increase...China is starting to take the market share...the overall market flat or slighly declining...nothing really new..I think that applies to most technologies today ;-)...Kris
You kinda stole my words, Kris. Everything seems to be as one would expect. I don't really see who the big Ethernet switch buyers WOULD be, if not data centers, and of course also ISPs and enterprise networks.
If enterpise networks outsource their data center functions, that doesn't sound odd either. Individual households, whenever possible, I'm pretty sure use WiFi rather than cabled Ethernet and switches. So aside from the one modem/router/switch combination box, I would expect any extra standalone Ethernet switches in households would be the exception.
Sorry @krisi, but I humbly disagree. 3D isn't dead either, but that is outside of the scope of this discussion. So you reckon we won't be at 4K in the next decade and a half (assuming that is how long it takes you to get to your next two TV's)?
I agree, 4k TV will happen in 15 years...but who cares?...it is a very increamental improvement that will take very long time to sell, spatial resolutionis limuted by your eyes afterall...we need to get a sense of smell and touch in the living room, that would be something interesting to write about
That shouldn't be a problem either. WiFi 802.11n goes up to 600 Mb/s in the 5 GHz band (although ours tops out at 270 Mb/s), which should be enough even for 4K with H.264 compression, and certainly enough for 4K with H.265 compression (which theoretically won't require much more bit rate than HD using H.262 compression, say on the order of 25 Mb/s for 60p). And now WiFi 802.11ad is moving up to the 60 GHz band, for up to 7 Gb/s capacity.
25G is a per channel/Lane protocol which is being driven by InfiniBand (EDR), Ethernet (100G 4x25) and Fibre Channel QSFP28. There is quite a bit of development work being done today across all 3 standards. It is anyone's guess what the adoption rate and cost will look like. This development willl likely fuel the need for many new intermediaries to allow 25G switch equipment to efficiently interop with existing servers and storage. Additionally, there are still many signal integrity challenges related to development and deployment of the 25G datacom architecture.
Rick, the interesting thing is the emergence of true 25Gbps fibers. 10GbE and 40GbE are carried on parallel fibers with usually a 2.5Gbps rate on each fiber. If you are building a 100,000 machine DC with 10GbE, that is a staggering amount of fiber.
So, when the transceiver story for 25Gbps settles down and they are shipping in volume at reasonable prices, there are some interesting questions. How many folks would be happy to take the speed boost to 25, use just one fiber, and say they are really happy? Win-win, what's not to like? Why make it more complicated by going to 50 (or 40, which would run fiber pairs at 20)? I suspect a lot of machinery can cruise just fine at 25 and never reach that ceiling. The supercomputer crowd, of course, will want 100.
The underlying technology is modular (and has been for some time) so we will see all kinds of variety in deployment, and it will not matter much. Just like 40 GbE and 10 GbE switches today use mostly the same components just configured differently.
The thing to watch is how the new 25Gbps transceivers reach the market, and the impact they are going to have no matter how they are configured.
We are not talking about long lines, that is a whole different market. The connections between ethernet switches in a DC do not use FDM or other expensive, physically large devices. Hence the long stall at 2.5Gbps since that was until recently the practical limit on how many bits a semiconductor could modulate or demodulate on one beam. Now Intel and partners are promising 25 Gbps devices.
Actually, multiple frequencies/colors would be just fine too, if you could meet the cost and packaging limits of an ethernet socket.
I haven't heard of 25Gig Ethernet, so I asume the 25Gig is for Inifiband, which is used with some storage switches (Infiniband also has 2.5Gig).
As mentioned, the other place you see 25G is 4x25Gig (100Gig, either 4 separate fibers or four mulitplexed frequencies), but I don't think that would show up in the chart as 25G.
The reason you don't see too much of 100Gig at one frequency (called coherent 100Gig), is cost. But, as with all things in the electronics industry, costs go down in time as they integrate more and can produce in volume.
A recent report from Dell Oro Group seems to contradict the findings of this report, stating that the market for Ethernet ports and services continues to grow. While the distribution of sales may be declining, I think the premise of the article is wrong. The premise is that with the cloud, Ethernet switch market sales will decline. While the efficiencies of the cloud could reduce the # of Ethernet ports sold versus an enterprise data center implementation, that doesn't consider the growing number of diverse cloud services enterprises are consuming. It also doesn't touch the desktop which has always dominated enterprise network purchase. Although the loss of many enterprise datacenters will be a reduction in some Ethernet port sales, those will be made up with the need for greater metro e service needs (higher port speeds) to GET TO and FROM the cloud.
My Mom the Radio Star Max MaxfieldPost a comment I've said it before and I'll say it again -- it's a funny old world when you come to think about it. Last Friday lunchtime, for example, I received an email from Tim Levell, the editor for ...
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...